TextClassificationSerial/EM.cpp at master · mremad/TextClassificationSerial · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
#include"EM.h"

EM::EM()
{

}

voidEM::classify_all_unlabeled_documents(NaiveBayesClassifier* classifier, int** feature_vectors, int number_unique_words, int number_documents, int number_labels)
{
for(int i = 0;i < number_documents;i++)
 {
int label = classifier->classify_unlabeled_document(feature_vectors[i], number_unique_words, number_labels);
 feature_vectors[i][0] = label;
 }
}

voidEM::copy_parameters(longdouble** src_likelihood, longdouble* src_prior, longdouble** dest_likelihood, longdouble* dest_prior, int number_labels, int number_unique_words)
{
for(int i = 0;i<number_labels;i++)
 {
for(int j = 0;j<number_unique_words;j++)
 {
 dest_likelihood[i][j] = src_likelihood[i][j];
 }

 dest_prior[i] = src_prior[i];
 }
}

boolEM::check_if_converged(longdouble** old_likelihood, longdouble* old_prior, longdouble** new_likelihood, longdouble* new_prior, int number_labels, int number_unique_words)
{
bool result = false;

longdouble total_diff = 0;

longdouble threshold = 0.0001;

for(int i = 0;i<number_labels;i++)
 {
for(int j = 0;j<number_unique_words;j++)
 {
 total_diff += fabs(old_likelihood[i][j]-new_likelihood[i][j]);
 }

 total_diff += fabs(old_prior[i]-new_prior[i]);
 }

printf("Total Difference: %Lf\n",total_diff);

if(total_diff < threshold)
 result = true;

return result;
}

voidEM::run_em(NaiveBayesClassifier* classifier, int** feature_vectors, int** labeled_docs, int** unlabeled_docs, int number_unique_words,int number_unlabeled_documents, int number_labeled_documents, int number_labels)
{

/*Initialize arrays to store old parameters*/

longdouble** old_likelihood = (longdouble**)malloc(sizeof(longdouble*)*number_labels);
longdouble* old_prior = (longdouble*)malloc(sizeof(longdouble)*number_labels);

for(int i = 0;i<number_labels;i++)
 old_likelihood[i] = (longdouble*)malloc(sizeof(longdouble)*number_unique_words);


/*Initial Step*/
//Construct classifier with labeled feature_vectors

 classifier->calculate_likelihood(labeled_docs, number_unique_words, number_labeled_documents, number_labels);

 classifier->calculate_prior(labeled_docs, number_labeled_documents, number_labels);

for(;;)
	{
break;
/*E Step*/

printf("Performing E Step\n");
classify_all_unlabeled_documents(classifier, unlabeled_docs, number_unique_words, number_unlabeled_documents, number_labels);

//ConsolePrint::print_2d_int(number_unique_words, number_unlabeled_documents, unlabeled_docs);
//ConsolePrint::print_2d_int(number_unique_words, number_labeled_documents, labeled_docs);
//ConsolePrint::print_2d_int(number_unique_words, number_labeled_documents+number_unlabeled_documents, feature_vectors);

/*M Step*/

printf("Performing M Step\n");
copy_parameters(classifier->get_likelihood(), classifier->get_prior(), old_likelihood, old_prior, number_labels, number_unique_words);

 classifier->calculate_likelihood(feature_vectors, number_unique_words, number_labeled_documents+number_unlabeled_documents, number_labels);

 classifier->calculate_prior(feature_vectors,number_labeled_documents+number_unlabeled_documents,number_labels);

/*Check for convergence*/

if(check_if_converged(old_likelihood,old_prior,classifier->get_likelihood(),classifier->get_prior(),number_labels,number_unique_words))
break;


	}

printf("EM Process Done\n");
}